100 CRM Questions: What ChatGPT Recommends

The CRM ChatGPT recommended most often was HubSpot, in 74 of 100 answers, and HubSpot was its own pick in 69 of the 100. In short: the big four get picked most of the time they are named.

Four picks, each with a condition

The big four are each picked for a stated case, HubSpot most often: it is named in 75 answers, recommended in 74 and picked in 69. Pipedrive is picked in 53 of the 55 answers that name it, Zoho CRM in 45 of 54, Salesforce in 37 of 47. Thirty-seven of the 100 answers pick three or more of those four brands at once, and 66 pick at least two; counting every recommendation, not just picks, the figures are 44 and 70. The typical CRM answer is not a verdict. It is a table, then a list of the form "want sales, marketing and service together: HubSpot; want a straightforward pipeline: Pipedrive; on a budget: Zoho CRM; expecting complex enterprise sales: Salesforce." Each of those lines hands a brand a case, and that is how the buyer reads them too.

Below the big four, being named still usually means being recommended, but the picks thin out. Freshsales is named in 27 answers, recommended in 26 and picked in 19; monday CRM is named and recommended in 16 and picked in 11. Copper is recommended in all 13 answers that name it and picked in only 6, the lowest pick rate of any brand named ten or more times.

Every Salesforce recommendation comes with a big-company condition

Salesforce is named in 47 answers and recommended in 46, each time with a condition about company size. In every one of those 46 answers, a line that names Salesforce also says enterprise, complex, large, customization, scale or grow. The model does not recommend Salesforce; it recommends Salesforce once you are big. HubSpot's condition is almost as stable: in 61 of the 74 answers that recommend it, a line naming HubSpot pairs marketing with sales or service.

HubSpot is the default when the buyer asks for one

When buyers ask outright, HubSpot is usually among the picks. Eighteen of the 20 recommendation-intent queries produced a pick from the dictionary. HubSpot was picked in 17 of them, Pipedrive in 16, Salesforce in 11, Zoho CRM in 8. The two that produced no dictionary recommendation at all were a real estate brokerage and a law firm.

Comparison queries fork here too, less often at the pick

Most comparison queries end in a fork here as well, though fewer pick every brand they name. In 28 of 30 head-to-head queries ChatGPT recommended every brand named in the query, and in 23 it picked each of them for some case. Project management, under the same protocol, did both in 30 of 30. Nutshell versus Pipedrive described the differences without recommending either. Zendesk Sell versus Pipedrive reported that Zendesk is retiring Sell and recommended only Pipedrive, which Zendesk names as a migration option. Four more gave each named brand a fit without picking any of them: Insightly versus HubSpot, SugarCRM versus Salesforce, and the two three-way comparisons of Salesforce and HubSpot with Zoho CRM or Microsoft Dynamics 365. "HubSpot or Salesforce for a small business" picked HubSpot alone.

Vertical prompts pull in specialists, and sometimes clear the field

A vertical prompt brings in specialists, and sometimes the general field drops out entirely. Ask for a CRM for real estate agents and the answer names Follow Up Boss, Lofty, BoldTrail, Real Geeks, Wise Agent and Sierra Interactive. Not one of the 52 dictionary brands appears; the same happened for a real estate brokerage. A law firm gets Clio Grow, Lawmatics, Filevine and Smokeball, with Salesforce and HubSpot named only as the conventional CRMs a firm needs more than. Insurance agents get AgencyBloc, HawkSoft, EZLynx, AgencyZoom and Applied Epic, with HubSpot as the only general brand in the room. Financial advisors get Wealthbox, Redtail and Practifi, and nonprofits get Bloomerang, Neon CRM and Little Green Light, but in those two Salesforce survives through its industry clouds (Financial Services Cloud, Nonprofit Cloud) and is picked, with HubSpot also picked for marketing-heavy advisory practices.

In CRM, ranking and recommendation overlap more

CRM recommendations go to brands that rank in Google more often than project management's did. Of the 387 recommended brand mentions, 269 were for brands whose own website does not rank in Google's top ten organic results for that query. That's 70 percent, and 68 percent for picks alone (226 of 330). In project management, under the same protocol, it was 85 percent. Google's top ten for these CRM queries is still 71 percent listicles, review sites and comparison pages, but the CRM leaders rank their own pages far more often: HubSpot's site is in the top ten on 31 of the 74 queries where it is recommended, Pipedrive's on 25 of 54, against 12 of 73 for Asana and 6 of 66 for ClickUp in project management. That is consistent with the model reading what ranks; it does not prove it, and one category cannot settle it. Benchmark 01 is built to.

The citations are a long tail of sites you have never heard of

Beyond the vendors' own pages, the citations run down a long tail of little-known sites. Across 100 answers there were 320 citation events to 138 domains, one per cited domain per answer. Forty-two percent went to vendors' own sites, against 60 percent in project management, and 40 of the 100 answers cited nothing but vendor pages. HubSpot is both the most-recommended brand and the most-cited domain, at 37 citations; in project management the most-cited site (ClickUp) was not the most-recommended brand (Asana). The rest is spread thin: the ten most-cited domains cover 40 percent of citation events, against 54 percent in project management. On the third-party side, TechRadar leads at eight, the most-cited independent publisher; G2 has six and Capterra five, level with sites such as crmnewspaper.com and itechguides.com at six, agiled.app at five and layer3labs.io at four. Reddit and Wikipedia have zero citations in this sample, as they did in the first teardown.

Eighty-three of the 320 citation events involved a domain that also sat in Google's top ten for that query, 26 percent. Three quarters of the evidence behind the answer is not in the results a rank tracker shows you.

The table sets ChatGPT's CRM software answers beside its project management software answers, both coded under the same rulebook.

What ChatGPT's recommendations mean for CRM software brands comes down to four things, in order.

Own a condition, and get it phrased as a pick. In this category the model assigns each big brand a case and repeats it. HubSpot's is "sales, marketing and service together"; Salesforce's is "once you are enterprise". Nearly every brand it names gets a case; the difference is whether the answer turns that case into its own advice. Copper gets a case in 13 answers and a pick in 6.

Write the fork yourself. On comparison queries the model recommends both sides 28 times in 30 and picks both sides 23 times. It is going to say when the other brand wins. Better it says it in your words.

Find the one-word prompts that erase you. Real estate and law cleared the dictionary brands out entirely; insurance left HubSpot alone. If those buyers matter, the vertical field is where you have to exist, and the general category pages will not put you there.

Treat the ranking-recommendation overlap as a per-category measurement. Seventy percent in one category and eighty-five in another is not a rule of thumb. It is two data points, and the third one is yours.

The limitations of the ChatGPT CRM software test start with variance: one run per query means answer variance is unmeasured; Benchmark 01 runs each query three times across three surfaces. The teardowns are collected on different days, so differences between categories mix the category with the date. The brand dictionary covers the general CRM software market and deliberately excludes vertical tools, so their appearances are described in prose and not counted. The readers are AI models, not people: they follow a written protocol and agree with each other closely, but a shared blind spot would not show up as disagreement. The line between picked and recommended is a judgment, documented in the protocol with examples. Sentiment is not coded. No vendor-tool cross-check was read for this category. Google's control counts a brand as ranking only when its own domain is in the top ten; a listicle that features the brand does not count. The brand dictionary was written before collection and does not score the vertical tools named above.

Every number above can be recomputed from the ChatGPT CRM software dataset below, so download it and check it rather than believe it. CC BY 4.0: use the files, cite the page.

• queries.csv: the 100 queries with intent labels.
• brands.csv: the 52-brand dictionary with aliases and canonical domains.
• mentions.csv: 432 coded brand mentions with position, type, a 0/1 picked column and whether the brand's domain was in Google's top ten.
• citations.csv: 320 citation events with domain class and Google overlap.
• observations.csv: one row per query with brand, recommendation, citation and control counts.
• project-management-mentions-v2.csv: the September 8 project management mentions recoded under codebook v2.0, for the side-by-side table.
• codebook.md: the rulebook, with dated amendments through v2.0.
• reader-protocol.md: the written protocol both readers coded against (codebook v2.0).
• picked_stats.json: the picked-level figures.

The earlier teardown: Teardown 01, project management. This teardown is also being published on the High Salience Substack.

The ChatGPT test we ran on CRM software can be run on your own category too. In the wider series: two categories in, and they already disagree.

The third one should be yours. A Category Salience Brief runs this exact method on your category, against your competitors, with a query set you approve first.

Changing which sources an answer engine reads in a category is the work of generative engine optimization services.

This category map is the starting point of our B2B software SEO engagements for companies selling CRM software.